Nguyen & Eger find AI bibliography age roughly halved in a decade and read it as field dynamism, not an NLP-specific citation bias

citation-analysisbibliometricscitation-amnesianlp-history

Nguyen & Eger (arXiv 2401.03545) study the age of works cited across AI subfields and report that the mean bibliography age has "roughly halved in the last 10 years" — references now reach back about half as far as they did a decade earlier. They interpret this as an artefact of the field's own pace of change (a fast-moving field has more recent work worth citing) rather than as evidence of an NLP-specific "citation amnesia" or a failure to engage prior literature.

The point this note holds is scope: Nguyen & Eger contest the age result, not the field-diversity result. Their dynamism reading pushes back on the claim that AI systematically forgets older work, but it says nothing against Wahle et al.'s finding that NLP's cross-field citation diversity collapsed, which is a measurement of breadth across fields, not depth back in time. A field can cite more recent work (age halving) while also citing a narrower range of fields (diversity falling); the two measurements are orthogonal, and the rebuttal lands only on the first.